Model the task before generating the item.
Know the concept, misconception, reasoning step, and difficulty driver first.
Launch in
The research basis for Classiqo.
Automated item generation, assessment engineering, expert review, and evaluation loops.
From cognitive model to classroom signal.
Keep professors in the loop where judgment matters.
Course alignment, validity, wording, and classroom fit need expert review.
Make quality measurable.
Attempts, difficulty, corrections, and edits should become product signals.
Papers that shape the product.
From item models to generated variants to real-world measurement.
01
The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring
Educational Researcher, 1984
tutoringlearning gainsmastery learning
Why it matters. Bloom reports an approximately 2-sigma tutoring effect, commonly interpreted as moving the average student from about the 50th percentile to about the 98th percentile.
Classiqo use. Sets the north star: tutor-like learning gains, but through scalable product workflows.
Read article02
AI Tutoring Outperforms Active Learning
Research Square preprint, 2024
AI tutoringHarvardeffect size
Why it matters. Harvard RCT evidence that structured AI tutoring can outperform active learning, with reported effect size around 0.63σ.
Classiqo use. Supports product design that couples pedagogy and tutor structure, not just raw model output.
Read article03
Developing a Taxonomy of Item Model Types to Promote Assessment Engineering
Journal of Technology, Learning, and Assessment, 2008
item modelsassessment engineeringIGOR
Why it matters. Item models make task variables explicit: stem, options, distractors, and supporting material.
Classiqo use. Professor feedback becomes structure: what can vary, what stays fixed, and what needs review.
Read article04
Using automatic item generation to create multiple-choice test items
Medical Education, 2012
AIGcognitive modelsmedical education
Why it matters. A practical three-stage AIG workflow: cognitive model, item model, generated items.
Classiqo use. Capture course reasoning, produce controlled candidates, keep professor review.
Read article05
Using a Hybrid of AI and Template-Based Method in Automatic Item
JMIR Formative Research, 2025
hybrid AIGLLMsexpert review
Why it matters. A bridge between template-based AIG and model-assisted authoring.
Classiqo use. Accelerate authoring while keeping checks, review states, and corrections visible.
Read article06
Assessing the Quality of AI-Generated Exams: A Large-Scale Field Study
arXiv, 2025
exam qualityIRTfield study
Why it matters. Connects generated questions to real course contexts and psychometric evaluation.
Classiqo use. Question quality needs measurement: item performance, review signals, reliability checks.
Read articleTrust comes from structure, review, and measurement.
Professors need structure, review, and response data. That is the product brief.